test driven development of scientific modelstest driven development of scienti c models tom clune...

Test Driven Development of Scientific Models

Tom Clune

Software Systems Support OfficeEarth Science Division

NASA Goddard Space Flight Center

May 1, 2012

Tom Clune (SSSO) TDD - NCAR May 1, 2012 1 / 61

Outline

1 Introduction

2 Testing

3 Testing Frameworks

4 Test-Driven Development

5 TDD and Scientific/Technical Software

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - Introduction - NCAR May 1, 2012 2 / 61

The Tightrope Act

Software development should not feel like this

The Tightrope Act

... or even like this

The Tightrope Act

Hopefully something more like this

The Development Cycle

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Extend

Compiles?

Executes?

Looks ok?

Correct?

Natural Time Scales

DesignEdit sourceCompilationBatchwaiting in queueExecutionAnalysis

Some observations

Risk grows with magnitude of implementation step

Magnitude of implementation step grows with cost ofverification/validation

Conclusion:Optimize productivity by reducing cost of verification!

Some observations

Risk grows with magnitude of implementation step

Magnitude of implementation step grows with cost ofverification/validation

Conclusion:Optimize productivity by reducing cost of verification!

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - Testing - NCAR May 1, 2012 7 / 61

Testing

Test Harness - work in safety

Collection of tests that constrain system

Detects unintended changes

Localizes defects

Improves developer confidence

Decreases risk from change

Localizes defects

Do you write legacy code?

“The main thing that distinguishes legacy code from non-legacy code istests, or rather a lack of tests.”

Michael FeathersWorking Effectively with Legacy Code

Lack of tests leads to fear of introducingsubtle bugs and/or changing thingsinadvertently.

Programming on a tightrope

This is also a barrier to involving pure

software engineers in the development of

our models.

Excuses, excuses ...

Takes too much time to write tests

Too difficult to maintain tests

It takes too long to run the tests

It is not my job

“Correct” behavior is unknown

http://java.dzone.com/articles/unit-test-excuses

- James Sugrue

It is not my job

- James Sugrue

It is not my job

- James Sugrue

It is not my job

- James Sugrue

It is not my job

- James Sugrue

It is not my job

- James Sugrue

Just what is a test anyway?

Tests can exist in many forms

Conditional termination:IF (PA( I , J)+PTOP.GT. 1 2 0 0 . ) &

c a l l s t o p m o d e l ( ’ADVECM: P r e s s u r e d i a g n o s t i c e r r o r ’ , 1 1 )

Diagnostic print statementp r i n t ∗ , ’ l o s s o f mass = ’ , d e l t a M a s s

Visualization of outputTemp1

5 10 15

Difference

5 10 15

Student Version of MATLAB

Analogy with Scientific Method?

Reality −→ RequirementsConstraints: theory and data −→ Constraints: testsFormulate hypothesis −→ Trial implementationPerform experiment −→ Run testsRefine hypothesis −→ Refine implementation

Properties of good tests

Isolating

I Test failure indicates location in source code

Orthogonal

I Each defect results in failure of small number of tests

Complete

I Each bit of functionality covered by at least one test

Independent

I No side effectsI Test order does not matterI Corollary: cannot terminate execution

Frugal

I Run quicklyI Small memory, etc.

Automated and repeatableClear intent

IsolatingI Test failure indicates location in source code

Orthogonal

I Each defect results in failure of small number of tests

Complete

Independent

Frugal

OrthogonalI Each defect results in failure of small number of tests

Complete

Independent

Frugal

CompleteI Each bit of functionality covered by at least one test

Independent

Frugal

IndependentI No side effectsI Test order does not matterI Corollary: cannot terminate execution

Frugal

FrugalI Run quicklyI Small memory, etc.

Automated and repeatable

Clear intent

Anatomy of a Software Test Procedure

testTrajectory() ! s = 12 at2

a = 2.; t = 3.

s = trajectory(a, t)

call assertEqual (9., s)call assertEqual (9., trajectory (2.,3.))

! no op

a = 2.; t = 3.

! no op

a = 2.; t = 3.

! no op

a = 2.; t = 3.

! no op

a = 2.; t = 3.

call assertEqual (9., s)

call assertEqual (9., trajectory (2.,3.))

! no op

a = 2.; t = 3.

! no op

a = 2.; t = 3.

! no op

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - Testing Frameworks - NCAR May 1, 2012 18 / 61

Testing Frameworks

Provide infrastructure to radically simplify:I Creating test routines (Test cases)I Running collections of tests (Test suites)I Summarizing results

Key feature is collection of assert methodsI Used to express expected results

c a l l a s s e r t E q u a l (1 20 , f a c t o r i a l ( 5 ) )

Generally specific to programming language (xUnit)I Java (JUnit)I Pnython (pyUnit)I C++ (cxxUnit, cppUnit)I Fortran (FRUIT, FUNIT, pFUnit)

GUI - JUnit in Eclipse

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - Test-Driven Development - NCAR May 1, 2012 22 / 61

(Somewhat) New Paradigm: TDD

Old paradigm:

Tests written by separate team (black box testing)

Tests written after implementation

Consequences:

Testing schedule compressed for release

Defects detected late in development ($$)

New paradigm

Developers write the tests (white box testing)

Tests written before production code

Enabled by emergence of strong unit testing frameworks

Old paradigm:

Consequences:

New paradigm

Old paradigm:

Consequences:

New paradigm

The TDD cycle

Benefits of TDD

High reliability

Excellent test coverage

Always “ready-to-ship”

Tests act as maintainable documentationI Test shows real use case scenarioI Test is maintained through TDD process

Less time spent debugging

Reduced stress / improved confidence

Productivity

Predictable schedule

Porting

Quality implementation?

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Benefits of TDD

High reliability

Productivity

Porting

Anecdotal Testimony

Many professional SEs are initially skepticalI High percentage refuse to go back to the old way after only a few days

of exposure.

Some projects drop bug tracking as unnecessary

Often difficult to sell to managementI “What? More lines of code?”

Not a panacea

Requires training, practice, and discipline

Need strong tools (framework + refactoring)

Does not invent new algorithms (e.g. FFT)I No such thing as magic

Maintaining tests difficult during a major re-engineering effort.

I But isnt the alternative is even worse?!!

Not a panacea

Does not invent new algorithms (e.g. FFT)

I No such thing as magic

Not a panacea

Maintaining tests difficult during a major re-engineering effort.I But isnt the alternative is even worse?!!

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - TDD and Scientific/Technical Software - NCAR May 1, 2012 28 / 61

The Challenge of Technical Software

Serious objections have been raised:

I Difficult to estimate errorF RoundoffF Truncation

I Stability/NonlinearityF Problems that occur only after long integrations

I Insufficient analytic casesI Test would just be re-expression of implementation

F Irreducible complexity?

These concerns largely revealI Lack of experience with software testingI Confusion between roles of verification vs validationI Burden of legacy software (long procedures; complex interfaces)

Serious objections have been raised:I Difficult to estimate error

F RoundoffF Truncation

I Insufficient analytic cases

I Test would just be re-expression of implementationF Irreducible complexity?

These concerns largely reveal

I Lack of experience with software testingI Confusion between roles of verification vs validationI Burden of legacy software (long procedures; complex interfaces)

These concerns largely revealI Lack of experience with software testing

I Confusion between roles of verification vs validationI Burden of legacy software (long procedures; complex interfaces)

These concerns largely revealI Lack of experience with software testingI Confusion between roles of verification vs validation

I Burden of legacy software (long procedures; complex interfaces)

Software Testing vs Science/Validation

Software tests should only check implementation.

Only a subset tests will express external requirements (i.e.implementation independent)

Other tests will reflect implementation choices

Use “convenient” input values - not realistic values

Consider tests for an ODE integrator implemented with RK4

A generic test may be for a constant flow field - any integrator shouldget an “exact” answer

A RK4 specific test may provide an artificial “flow field” that returnsthe values 1.,2.,3.,4. on subsequent calls independent of thecoordinates

Test by Layers

Do test

Proper # of iterations

Pieces called in correct order

Passing of data betweencomponents

Do NOT test

Calculations inside components

Much easier to do in practice with objects than with procedures.

Numerical Tolerance

For testing numerical results, a good estimate for the tolerance isnecessary:

If the tolerance is too low, then the test may fail for uninterestingreasons.

If the tolerance is too high, then the test may have no teeth

Unfortunately ...

Error estimates are seldom available for complex algorithms

And of those, usually we just have an asymtotic form with unknownleading coefficient!

Numerical Tolerance

Unfortunately ...

Numerical Tolerance

Unfortunately ...

Numerical Tolerance

Unfortunately ...

Numerical Tolerance

Unfortunately ...

Numerical tolerance (cont’d)

Observations

1 machine epsilon is a good estimate for most short arithmeticexpressions

2 large errors arise in small expressions in fairly obvious places (1/∆)

3 larger errors are generally a result of composition of many operations

Conclusion: If we write software as a composition of distinct smallfunctions and subroutines, the errors can be reasonably bounded at eachstage

Observations

TDD and long integration

TDD does not directly relate to issues of stability

If long integration gets incorrect results:

1 Software defect: missing test2 Genuine science challenge

TDD can reduce the frequency at which long integrations areneeded/performed

If long integration gets incorrect results:1 Software defect: missing test

2 Genuine science challenge

If long integration gets incorrect results:1 Software defect: missing test2 Genuine science challenge

TDD and Lack of Analytic Results

Keep in mind: “How can you implement it if you cannot say what itshould do?”

Split into pieces - often each step has analytic solution

Choose input values that are convenient

Consider a trivial case:

c a l l a s s e r t E q u a l ( 3 . 1 4 1 5 9 2 6 5 , a r e a O f C i r c l e ( 1 . ) )c a l l a s s e r t E q u a l ( 6 . 2 8 . . . , a r e a O f C i r c l e ( 2 . ) )

What if instead the areaOfCircle() function accepted 2 arguments: “π”and r .

c a l l a s s e r t E q u a l ( 1 . , a r e a O f C i r c l e ( 1 . , 1 . ) )c a l l a s s e r t E q u a l ( 4 . , a r e a O f C i r c l e ( 1 . , 2 . ) )c a l l a s s e r t E q u a l ( 2 . , a r e a O f C i r c l e ( 2 . , 1 . ) )

TDD and irreducible complexity

Are the tests as complex as the implementation?

Short answer: No

Long answer: Well, they shouldn’t be ...I Unit tests use specific inputs - implementation handles generic caseI Each layer of algorithm is tested separatelyI Layers of the production code are coupled - huge complexityI Tests are decoupled - low complexity

Short answer: No

Long answer: Well, they shouldn’t be ...

I Unit tests use specific inputs - implementation handles generic caseI Each layer of algorithm is tested separatelyI Layers of the production code are coupled - huge complexityI Tests are decoupled - low complexity

Short answer: No

Long answer: Well, they shouldn’t be ...I Unit tests use specific inputs - implementation handles generic case

I Each layer of algorithm is tested separatelyI Layers of the production code are coupled - huge complexityI Tests are decoupled - low complexity

Short answer: No

Long answer: Well, they shouldn’t be ...I Unit tests use specific inputs - implementation handles generic caseI Each layer of algorithm is tested separately

I Layers of the production code are coupled - huge complexityI Tests are decoupled - low complexity

Short answer: No

Long answer: Well, they shouldn’t be ...I Unit tests use specific inputs - implementation handles generic caseI Each layer of algorithm is tested separatelyI Layers of the production code are coupled - huge complexity

I Tests are decoupled - low complexity

Short answer: No

Long answer: Well, they shouldn’t be ...I Unit tests use specific inputs - implementation handles generic caseI Each layer of algorithm is tested separatelyI Layers of the production code are coupled - huge complexityI Tests are decoupled - low complexity

TDD and the Legacy Burden

TDD was created for developing new code, and does not directlyspeak to maintaining legacy code.

Adding new functionalityI Avoid wedging new loging directly into existing large procedureI Use TDD to develop separate facility for new computationI Just call the new procedure from the large legacy procedure

RefactoringI Use unit tests to constrain existing behaviorI Very difficult for large proceduresI Try to find small pieces to pull out into new procedures

TDD Best Practices

Small steps - each iteration � 10 minutes

Small, readable tests

Extremely fast execution - 1 ms/test or less

Ruthless refactoring

Verify that each test initially fails

TDD Best Practices

TDD and Performance

Optimized algorithms may require many steps within a singleprocedure

TDD emphasizes small simple procedures

Such an approach may lead to slow execution

Solution: BootstrappingI Use initial solution as unit test for optimized solutionI Maintain both implementations

Experience to date

TDD has been used heavily within several projects at NASA

Mostly for “infrastructure” portions - relatively little numerical alg.

pFUnit

DYNAMO - spectral MHD code on shperical shell

GTRAJ - offline trajectory integration (C++)

Snowfake - virtual snowfakes; Multi-lattice Snowfake

Observations:

∼ 1:1 ratio of test code to source code

Works very well for infrastructure

Learning curveI 1-2 days for techniqueI Weeks-months to wean old habitsI Full benefit may require some sophistication

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - Example - NCAR May 1, 2012 42 / 61

Linear Interpolation

��

-xx1 x2 x3 xnx

t tt t

Potential Tests

Bracketing: Find i such that xi <= x < xi+1

Computing node weights:

wa =xi+1 − x

xi+1 − xiwb = 1− wa

Compute weighted sum: y = waf (xi ) + wbf (xi+1)

Potential Tests

wa =xi+1 − x

Potential Tests

wa =xi+1 − x

Potential Tests

wa =xi+1 − x

Bracketing Tests

i n d e x = b r a c k e t ( nodes , x )

Case Preconditions Postconditionnodes x return

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1

other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2

at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)

at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)

other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)

out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds error

out-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Bracketing Tests

interior {x} = {1, 2, 3} x = 1.5 i = 1other interior {x} = {1, 2, 3} x = 2.5 i = 2at node {x} = {1, 2, 3} x = 2.0 i = 2 (?)at edge {x} = {1, 2, 3} x = 1.0 i = 1 (?)other edge {x} = {1, 2, 3} x = 3.0 i = 2 (????)out-of-bounds {x} = {1, 2, 3} x = 1.5 out-of-bounds errorout-of-order {x} = {1, 2, 3} x = 1.5 out-of-order error

Example: Bracketing Test 1

Preconditions: {x} = {1, 2, 3}, x = 1.5

Postcondition: return 1

s u b r o u t i n e t e s t B r a c k e t 1 ( )nodes = [ 1 . , 2 . , 3 . ]i n d e x = g e t B r a c k e t ( nodes , 1 . 5 )c a l l a s s e r t E q u a l ( 1 , i n d e x )

end s u b r o u t i n e

s u b r o u t i n e t e s t B r a c k e t 1 ( )c a l l a s s e r t E q u a l ( 1 , g e t B r a c k e t ( [ 1 . , 2 . , 3 . ] , 1 . 5 ) )

f u n c t i o n g e t B r a c k e t ( nodes , x ) r e s u l t ( i n d e x )i n d e x = 1

end f u n c t i o n

f u n c t i o n g e t B r a c k e t ( nodes , x ) r e s u l t ( i n d e x )i f ( x > nodes ( 2 ) ) then

i n d e x = 2e l s e

i n d e x = 1end i f

end f u n c t i o n

Generalize ...

f u n c t i o n g e t B r a c k e t ( nodes , x ) r e s u l t ( i n d e x )

do i = 1 , s i z e ( nodes ) 1i f ( nodes ( i +1) > x ) i n d e x = i

end do

end f u n c t i o n

Generalize ...

end do

end f u n c t i o n

Generalize ...

end do

end f u n c t i o n

Generalize ...

end do

end f u n c t i o n

Tests for Computing Weights

Case Preconditions Postconditioninterval x weights

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]

upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]

interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]

big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]

degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate error

out-of-bounds [1., 2.] x = 0.5 out-of-bounds error

lower bound [1., 2.] x = 1.0 w = [1.0, 0.0]upper bound [1., 2.] x = 1.0 w = [0.0, 1.0]interior [1., 2.] x = 1.5 w = [0.5, 0.5]big interval slope [1., 3.] x = 1.5 w = [0.75, 0.25]degenerate [1., 1.] x = 1.0 degenerate errorout-of-bounds [1., 2.] x = 0.5 out-of-bounds error

Example: Weights Test 1

Precondition: [a, b] = [1., 2.], x = 1.0

Postcondition: w = {1.0, 0.0}

s u b r o u t i n e t e s t W e i g h t 1 ( )r e a l : : i n t e r v a l ( 2 ) , w e i g h t s ( 2 )r e a l : : xi n t e r v a l = [ 1 . , 2 . ]w e i g h t s = computeWeights ( i n t e r v a l , 1 . 0 )c a l l a s s e r t E q u a l ( [ 1 . 0 , 0 . 0 ] , w e i g h t s )

end s u b r o u t i n e t e s t W e i g h t 1

r e a l f u n c t i o n computeWeights ( i n t e r v a l , x ) r e s u l t ( w e i g h t s )r e a l , i n t e n t ( i n ) : : i n t e r v a l ( 2 )r e a l , i n t e n t ( i n ) : : xw e i g h t s = [ 1 . 0 , 0 . 0 ]

end f u n c t i o n

Example: Tying it together

Precondition:I {(x , y)i} = {(1, 1), (2, 1), (4, 1)}I x = 3

Postcondition: y = 1.

s u b r o u t i n e t e s t I n t e r p o l a t e C o n s t a n t Y ( )r e a l : : nodes ( 2 , 3 )nodes = r e s h a p e ( [ [ 1 , 1 ] , [ 2 , 1 ] , [ 4 , 1 ] ] , shape = [ 2 , 3 ] )c a l l a s s e r t E q u a l ( 1 . 0 , i n t e r p o l a t e ( nodes , 3 . 0 ) )

end s u b r o u t i n e t e s t I n t e r p o l a t e 1

f u n c t i o n i n t e r p o l a t e ( nodes , x )r e a l , i n t e n t ( i n ) : : nodes ( : , : )y = 1

end f u n c t i o n i n t e r p o l a t e

Example: Tying it together

Precondition:I {(x , y)i} = {(1, 1), (2, 3), (4, 1)}I x = 3

Postcondition: y = 2.

s u b r o u t i n e t e s t I n t e r p o l a t e 1 ( )r e a l : : nodes ( 2 , 3 )nodes = r e s h a p e ( [ [ 1 , 1 ] , [ 2 , 3 ] , [ 4 , 1 ] ] , shape = [ 2 , 3 ] )c a l l a s s e r t E q u a l ( 1 . 0 , i n t e r p o l a t e ( nodes , 3 . 0 ) )

end s u b r o u t i n e t e s t I n t e r p o l a t e 1

f u n c t i o n i n t e r p o l a t e ( nodes , x ) r e s u l t ( y )i n t e g e r : : ir e a l : : w e i g h t s ( 2 ) , xAtEndPoints ( 2 ) , y A t E n d p o i n t s ( 2 )

i = g e t B r a c k e t ( nodes ( 1 , : ) , x )

xAtEndPoints = nodes ( 1 , i ) ! used d e r i v e d type ?y A t E n d p o i n t s = nodes ( 2 , i )w e i g h t s = computeWeights ( nodes ( 1 , [ i , i +1]) , x )

y = sum ( w e i g h t s ∗ y A t E n d p o i n t s )end f u n c t i o n i n t e r p o l a t e

Outline

1 Introduction

2 Testing

6 Example

7 pFUnit

Tom Clune (SSSO) TDD - pFUnit - NCAR May 1, 2012 59 / 61

pFUnit - Fortran Unit testing framework

Tests written in Fortran

Supports testing of parallel (MPI) algorithms

Support for multi-dimensional array assertions

Written in standard F95 (plus a tiny bit of F2003)

Developed using TDD

Tutorial in the afternoon sessioon

References

pFUnit: http://sourceforge.net/projects/pfunit/

Tutorial materialsI https://modelingguru.nasa.gov/docs/DOC-1982I https://modelingguru.nasa.gov/docs/DOC-1983I https://modelingguru.nasa.gov/docs/DOC-1984

TDD Bloghttps://modelingguru.nasa.gov/blogs/modelingwithtdd

Test-Driven Development: By Example - Kent Beck

Mller and Padberg,”About the Return on Investment of Test-DrivenDevelopment,” http://www.ipd.uka.de/mitarbeiter/muellerm/

publications/edser03.pdf

Refactoring: Improving the Design of Existing Code - Martin Fowler

JUnit http://junit.sourceforge.net/

These slides https://modelingguru.nasa.gov/docs/DOC-2222

test driven development of scientific modelstest driven development of scienti c models tom clune...

Documents